MIT Researchers are reporting that they have created a system for self assembling robots… almost.
In a cute video they show a collection of colored cubes bouncing around and latching onto each other in various configurations. They have solved an enormous number of interesting engineering problems, and no doubt there are some applications (a Rubik cube that can solve itself would get my money), but… are they really self assembling robots?
They assemble themselves into configurations on the basis of the minimal unit of one cube. They have no way of building a cube themselves, and hence replacing one that is broken. Perhaps that is too harsh a requirement to qualify as self-assembling: after all assembly means only putting the parts together and these guys are defined such that the parts are the individual cubes.
More importantly if you listen carefully to the researchers they reveal that all the software runs on external computers and communicates with the cubes to make them jump into the required configurations. In this sense they are not at all self assembling, the external device has an overview of the system and determines the next configuration. Thus the end product is not an emergent organisation from a set of decentralized autonomous units, but good old fashioned centralized control.
Sigh, it seems true self-assembly will have to wait for another day.
Last month the non-government education research outfit, ACER, issued a media release that was merely the latest in pretty much an unbroken string of bad news for Australian school quality.
“A comprehensive new analysis from the Australian Council for Educational Research (ACER) has provided evidence beyond simple international rankings that the achievement levels of Australian students declined in the period 2000 to 2012.”
In more detail they found:
• an overall decline in the reading and mathematics levels of Australian 15-year-olds
• variation in the decline in the reading levels of Australian 15-year-olds, with greater declines in Tasmania, South Australia, New South Wales and the ACT than in Western Australia, the Northern Territory, Victoria and Queensland
• a small improvement in Year 4 mathematics levels since 1994 and a small improvement in Year 3 reading levels from 2008 to 2012, and
• a growing gap between the most advantaged and the least advantaged secondary schools in Australia.
This is on top of the annual international rankings, TIMSS and PISA, on which Australia continues to slide. This is a well known problem. Andrew Leigh, before he entered politics, wrote about it and concluded that there had been “a fall in school productivity.”
ACER says more funding is needed. That was Gillard’s solution too, which is why she brought in Gonski. But lack of funding is not the problem. Australian schools did fine half a century ago with much smaller budgets.
The reason is simply that the education system has been stripped of content in the core areas of numeracy and literacy. They just don’t teach as much of it. Find yourself a maths curriculum, or a literacy curriculum, from forty years ago and see how much more was being taught in Australian schools.
Take my local primary school as a simple example. The kids do a maths class only three days a week. And those classes rarely involve practice, exercises, or times tables drills, because that’s boring and not fun and “we don’t teach that way any more.”
The effect of protecting the children from the drudgery of skills-based approach to maths and literacy can be seen in the following chart (from an ACER report):
Notice that the East Asian countries are outperforming us, Europe and North America. And by a lot. The difference is huge. Note also the percent of students who are in the “advanced” category. Those people are rare as hen’s teeth in Sweden (3 percent) but make up nearly half the students in a typical school in Singapore. We’re at 10 percent. This fits with the anecdotal evidence from lecturers on Australian campuses. University teachers in numeracy based subjects despair at the small numbers, in historical terms, of students who are highly numerate.
Starting about three decades ago, the Western education bureaucracies became overrun with ideologues. These activists had little interest empirical studies that investigate which techniques work, and which don’t. They firmly believed that drills and memorization of any kind were bad. They stripped back the maths that girls struggled with. They introduced a verbal component to maths (such as “explaining your working” – see here for a striking example ). Since boys tend to have better spatial ability and girls tend to have better verbal ability, taking out spatial skills and putting more “verbal ability” into maths tips the scales towards girls. It helps girls bridge the maths gap not by helping girls do maths better but by redefining the subject itself.
The ideological takeover that has caused the decline in standards presents a problem for any review of teaching methods, because most of the local ‘experts’ in Australia are invested in the current, failing system. (I say ‘most’…. it’s not quite all). But as the chart shows, this is across the Western world, driven by a de-emphasizing of domain specific knowledge, practice, facts, and core material; and an emphasis on learning that is more fun, self-directed, high level and cultural. The nations that are doing it right are the East Asian nations such as Korea, Hong Kong, Singapore. The proof of the pudding is in the eating. They’re feasting right now and we are starving.
Therefore there is a simple solution. The TIMSS and PISA rankings tell us which countries are doing best in numeracy and literacy. In both cases the best performing Western nation is South Korea. We could import the South Korean curriculum and teaching methods in toto, and work from there. But Singapore is not far behind Korea, and is far ahead of us. Singapore would possibly be a better choice as there is a common language and shared cultural history.
This would of course mean that numerous committees, organizations, and advisors would become instantly redundant but they could certainly reskill to cope with the new, higher quality education system. It would also be a shock to those who think that the West should lecture and teach other nations (about, for example, democracy and global warming) but that we have nothing to learn from them.
If we swallow our pride, Asia can help us fix our broken education system.
From slashdot comes news of a scientist who can grow meat that is perfectly edible from a ‘culture’, in other words, in a lab.
“Today, at 14:00 Western European Time (9:00 am Eastern), Professor Mark Post of Maastricht University ( the Netherlands ) will present a world first: he will cook and serve a burger made from Cultured Beef in front of an invited audience in London. The event will include a brief explanation of the science behind the burger. You can witch the event live, online. The project’s fact sheet is to be found here (pdf).”
It’s real meat, too, not some look-alike. These are animal muscle cells that are encouraged to grow and multiply, just as they would if they were part of a complete living organism. HEre are some of the steps involved, as described by the project’s fact sheet from Maastricht University.
The first step is to extract muscle stem cells from animals, usually cows, pigs or chickens. This project uses stem cells obtained from little pieces of fresh cow muscle for instance obtained through biopsy.
– The cells must then multiply, which requires a growth medium.
This could create a win-win situation for vegetarians where they can enjoy the benefits of meat knowing that no animal was harmed in the making of it. It also creates the possibility of synthesizing the meat of rare animals and creating fresh food in space.
When some plants are attacked by aphids they can defend themselves by releasing ‘volatiles’ into the air, chemicals that repel aphids and also attract wasps, which are an enemy of aphids. If plants could tell each other when aphids are attacking, they could release their volatiles earlier and thwart the attack. It turns out that they might be doing just that.
But how are they communicating? It’s not quite clear yet but it appears to be through a fungal network that infiltrates plant roots. Therefore, the fungi might be operating as a diffuse multi-organism nervous system.
Boston Dynamics, in partnership with DARPA, have developed Atlas – a humanoid robot that will be used for ‘friendly’ tasks such as disaster response.
Atlas has been developed for the Robotics Challenge, a DARPA funded competition in which robots must complete a series of difficult tasks. DARPA runs the Robotics Challenge in the hope that it will promote breakthrough developments in robotics that lead to the development of intelligent machines for disaster response activities.
Despite the benign, harmless goals of both the challenge and Atlas, some have suggested that the robot heralds the beginning of sophisticated killer robots such as those seen in the Terminator movies. While this might sound outlandish, it is quite possible that a robot that could traverse rough terrain and undertake complex tasks on its own could also be used for military purposes. And indeed DARPA’s core mission is not humanitarian, its purpose is to advance military technology. Once the technology is born, it breeds new generations and new variations as other researchers and engineers adapt the breakthrough to novel purposes.
So it’s true that Atlas is harmless.
We can’t be so confident about Atlas’ children.
Picture is copyright DARPA. More on Atlas:
Pictures of Atlas here.
Slate: DARPA’s new robot looks like the Terminator, or maybe Cleatus
Boston Dynamics: Atlas: The Agile Robot
New York Times: Modest debut of Atlas may foreshadow age of ‘Robo Sapiens’
Cnet: Domo Arigato, Mr Atlas; Be Afraid: DARPA unveils terminator-like Atlas robot
The Australian: Atlas is not a terminator, it’s a helpful robot, says US Army
As it gets better and better it will be less like people. It’s going to be its own kind of intelligence.
-Gary Marcus, NYU
When will AI get to the point where it is as powerful as a human brain? Or when it can mimic a human mind perfectly?
That’s called “strong AI” and it is a long way off. In some ways computer minds are already stronger than human minds: they can beat us at chess, calculate staggering mathematical solutions, pilot planes, track targets and many more tasks better than we can. But these are specialized domains and a computer program cannot yet successfully emulate a human in all its abilities and faculties.
Even so, computers are able to do more and more tasks for us, often tasks that we were never able to do ourselves. Will they become conscious? We can’t answer that because we don’t have a clear understanding of what consciousness is ourselves.
There is a short (8 minute) documentary on the present and future of AI and robots from PBS.
(More commentary here)